How many times have you tried to capture the essence of a particular chat window, to pass through to another one? With a fresh context window?
Or do you feel like you have ideas and projects and insights scattered across thirty different chats, notes, and files?
It’s not a tooling problem. It’s a biological one.
The Old Guard
Cal Newport wrote a book about this. So did Greg McKeown.
Newport’s primary enemy was the screen. Social media. Surveillance capitalism. The slot machine it required to keep running.
His solution was simple:
Go deep, not wide. Shut out the noise. Protect the signal.
McKeown’s enemy was the inability to say no. To recognise what was essential and what was not.
His solution was even simpler:
Do fewer things, better.
They were good books. But they’re useless now.
Because the enemy has changed.
It is no longer just the noise coming at you, (…though there’s more of that than ever before).
It is the noise coming from you.
The Flood
I reckon most people have generated more text, more images, more reports in the last twelve months than they would have read (not even written, just read) in their entire working life before AI.
Think about that for a moment.
Before the flood, a marketing team might produce one campaign brief a week.
One deck. One set of ad variants. The constraint was production. The bottleneck was the keyboard.
AI removed that bottleneck overnight.
Now the same team generates fifteen variations of a brief before lunch.
Within each of those briefs are now forty ad concepts, twelve landing pages, and a full content calendar.
They feel fast. They feel productive.
But the work still has to be read.
It still has to be judged. It still has to be held up to the light by a human brain that processes language at the same speed it did when they were writing those briefs by hand.
The throughput went up a thousand fold. The processing didn’t.
Biology
The issue is not the speed.
It’s the immutable fact that your brain has a context window.
It’s ~40-60 bits per second of conscious processing. (I googled it. We’ve measured it.)
And it has been that number for as long as we’ve been measuring it.
It was that number when you were reviewing one brief a week. It is still that number now that you are reviewing fifteen before lunch.
You cannot upgrade it. You cannot scale it. You cannot throw more compute at it.
Every piece of AI-generated output that crosses your desk makes a withdrawal from that budget.
Every draft you skim instead of read. Every approval you absently grant. Every “I’ll get to it” you utter.
Three hours behind turns into three tasks off track, because you’re running with three percent capacity.
The quality assurance step did not disappear. The packaging step did not disappear.
And if it did, congratulations:
You’re now the equivalent of a human ChatGPT wrapper.
Bandwidth
Models get faster, context windows get longer, costs get cheaper. That curve will continue.
The bottleneck that will never improve is you.
Your ability to judge what is right. To catch what is wrong. To decide what ships and what doesn’t.
To look at a thousand ideas and know, really know, which three are correct and which nine hundred and fifty are plausible garbage.
That is not a skill you can automate. It is the very thing that makes you valuable.
Because you are not running out of output. You are running out of the one thing that makes the output worth anything:
You are running out of bandwidth.
Newport told you to protect your attention from the world. McKeown told you to protect your time from other people. The next book needs to tell you to protect your bandwidth from yourself.
I could even “write” it tomorrow, if you like.
But you wouldn’t read it. Neither would I.
I tried to write that story a few weeks later. It came out differently than I expected.